[ES|QL] Force ES client timeout after 2 minutes (#168929)

## Summary

We found out that for ES|QL queries while the kibana server (and
browser) timeouts in 2 minutes, the requests are still running in ES. So
instead of being aborted in 2 minutes, the hit a 5minute timeout from
the proxy which retries 3 times and then aborts.

This happens ONLY when bfetch is enabled (which is the default)

After an investigation with Rudolf it seems that we don't abort
correctly in bsearch. Lukas is fixing it here
#169041 .

src/plugins/data/server/search/strategies/esql_search/esql_search_strategy.ts

@@ -11,6 +11,8 @@ import type { Logger } from '@kbn/core/server';
 import { getKbnServerError, KbnServerError } from '@kbn/kibana-utils-plugin/server';
 import type { ISearchStrategy } from '../../types';
 
+const ES_TIMEOUT_IN_MS = 120000;
+
 export const esqlSearchStrategyProvider = (
   logger: Logger,
   useInternalUser: boolean = false
@@ -23,6 +25,17 @@ export const esqlSearchStrategyProvider = (
    * @returns `Observable<IEsSearchResponse<any>>`
    */
   search: (request, { abortSignal, ...options }, { esClient, uiSettingsClient }) => {
+    const abortController = new AbortController();
+    // We found out that there are cases where we are not aborting correctly
+    // For this reasons we want to manually cancel he abort signal after 2 mins
+
+    abortSignal?.addEventListener('abort', () => {
+      abortController.abort();
+    });
+
+    // Also abort after two mins
+    setTimeout(() => abortController.abort(), ES_TIMEOUT_IN_MS);
+
     // Only default index pattern type is supported here.
     // See ese for other type support.
     if (request.indexType) {
@@ -41,8 +54,10 @@ export const esqlSearchStrategyProvider = (
             },
           },
           {
-            signal: abortSignal,
+            signal: abortController.signal,
             meta: true,
+            // we don't want the ES client to retry (default value is 3)
+            maxRetries: 0,
           }
         );
         return {